attributes serialization

[MarcAntoine-Arnaud]

How to serialise attributes on XML items ?
I d'ont see anything for that. Any suggestion ?

[oli-obk]

Yea attributes are a weird thing. I'm not sure how to handle them. We could rename fields to something like %foo, where the % denotes that the field is meant to be an attribute. This would require one to annotate all attributes, but would work quite nicely I think and remove all confusion we have currently in the deserializer.

[MarcAntoine-Arnaud]

Yep I think like in every library who manage XML ... It's still the same problem of naming.
But yep adding a prefix can made sense (which need to be a parameter) because with that we can generate reversible JSON.

[RReverser]

This would require one to annotate all attributes, but would work quite nicely I think and remove all confusion we have currently in the deserializer.

I agree, also was thinking about this approach to ensure roundtrips. Could avoid most annotations by using something like attr_width vs sub_item by default, but maybe that will be even more verbose?

[oli-obk]

Could avoid most annotations by using something like attr_width vs sub_item by default, but maybe that will be even more verbose?

You mean to require adding % for non-attributes?

[RReverser]

@oli-obk I mean to avoid any annotations and instead use prefixes for struct fields (which is essentially the same, just perhaps more convenient).

[oli-obk]

hmm.. I don't like that, because it exposes xml stuff to users of the struct who don't even care about xml

[RReverser]

Maybe... although author can still add annotations to rename it if it's a public structure / fields and not just for internal usage.

[ghost]

Probably you already know of it, but here there's a nice overview of some of the most common mapping conventions for attributes (and namespaces). The main problem is, there's all kind of conventions. Maybe a plugin system to allow choosing one's preferred method would be best?

[Boscop]

I think an annotation would be a good choice. Maybe #[xml(attribute)].

[oli-obk]

I think an annotation would be a good choice. Maybe #[xml(attribute)].

Serde has no support for such features. serde-xml is just a serde format, we can't create features that serde doesn't have and we can't parse extra attributes.

[oli-obk]

We'll also need to enforce that attribute-fields come before regular fields, because the serializer cannot look ahead. This is probably only doable with a runtime error though.

[anton-dutov]

So, may be like $value, but more complex #[serde(rename = "$xml:attr:name")] #[serde(rename = "$xml:value")]

[Boscop]

@anton-dutov Wouldn't that also affect json serializaton because it's #[serde(rename)]?

[anton-dutov]

@Boscop Just for case where only xml (seri/deseri)lization used, on other cases modification of serde required

#[rename(field, target=json)]
#[rename(Field, target=xml)]
field: Type

[MarcAntoine-Arnaud]

it's interesting @anton-dutov, in this formulation #[serde(rename = "$xml:attr:name")] we can also address the namespace like: #[serde(rename = "$xml:ns:name")] (issue #50).
On my side it looks the most relevant solution.

[anton-dutov]

@MarcAntoine-Arnaud, @Boscop

Continue discussion in #62

[RinChanNOWWW]

Hello guys, is this feature supported now?

[ThomasMcandrew]

For anyone here in the future, this is implemented at least I am on 0.6.0

#[derive(Serialize, Deserialize)]
pub struct Layout {
    #[serde(rename = "control")]
    pub control: Vec<Control>,

    #[serde(rename = "@height")]
    pub height: String,

    #[serde(rename = "@width")]
    pub width: String,
}

Anything renamed with the @ will map to an attribute, otherwise they will be treated as a child component.

[qsantos]

Note that the @ is only for serialization. You should not use it for deserialization. Extending the example above:

use serde::{Deserialize, Serialize};
use serde_xml_rs::{from_str, to_string};

#[derive(Debug, Serialize, Deserialize)]
pub struct Control {
    name: String,
}

#[derive(Debug, Serialize, Deserialize)]
pub struct Layout {
    #[serde(rename(serialize = "@height"))]
    pub height: String,

    #[serde(rename(serialize = "@width"))]
    pub width: String,

    #[serde(rename = "control")]
    pub control: Control,
}

fn main() {
    let xml = r#"
        <Layout height="3" width="5">
            <control>
                <name>Hello World!</name>
            </control>
        </Layout>
    "#;

    let layout: Layout = from_str(xml).unwrap();

    println!("{layout:#?}");
    println!("{}", to_string(&layout).unwrap());
}

will output:

Layout {
    height: "3",
    width: "5",
    control: Control {
        name: "Hello World!",
    },
}
<?xml version="1.0" encoding="UTF-8"?><Layout height="3" width="5"><control><name>Hello World!</name></control></Layout>

[kitplummer]

Is there a way to serialize a HashMap to a series of Attributes?

I can get the attributes deserialized into a HashMap.

<detail><_flow-tags_ ASV_h91="2024-04-19T00:36:52Z" Ditto="2024-04-19T00:40:52Z" /></detail>

But, no clue how to serialize that. Here's what I have in code:

#[derive(Debug, serde::Serialize, Deserialize, PartialEq)]
pub struct Detail {
    #[serde(rename(deserialize = "_flow-tags_", serialize = "@_flow-tags_"))]
    pub flow_tags: HashMap<String, String>,
}

Any ideas? I'm guessing it is actually not implemented. :)
I get the following error:

panicked at /Users/kitplummer/.cargo/registry/src/index.crates.io-6f17d22bba15001f/serde-xml-rs-0.6.0/src/ser/plain.rs:180:9:
not implemented
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Kinda curious why it is trying to handle a PlainStringSerializer.